Smoothing compositional gradient and prox-linear approximate gradient methods reach stationary points in O(1/(δε²)) and O(1/ε²) iterations for structured non-smooth non-convex compositions.
Conservative set valued fi elds, automatic differentiation, stochastic gradient method and deep learning, 2020
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Note on Complexity for Two Classes of Structured Non-Smooth Non-Convex Compositional Optimization
Smoothing compositional gradient and prox-linear approximate gradient methods reach stationary points in O(1/(δε²)) and O(1/ε²) iterations for structured non-smooth non-convex compositions.